An instrument that cannot register a change reports “no change” — the same words a working instrument uses for a real result. The two are indistinguishable from the outside, which is what makes this failure survive review: nothing looks wrong.
So the question a number owes an answer to is not “is this right?” but “what would have made this different?” If nothing would have, the number is decoration.
The practical form is a control. Perturb the thing the measurement should detect and confirm the reading moves; assert the instrument observed something before believing what it observed. Both are cheap, and both have caught conclusions that had already been reported as findings.
The same applies to a guard: one that cannot fail for the right reason is a comment with a test runner attached.